The research proposes a new methodological framework based on dynamic priority to handle different order classes in robotic warehouse systems. Traditional static priority methods in facility logistics may cause low-priority orders to experience excessive delays and fail to ensure fairness. Our dynamic priority approach addresses this fairness issue by adjusting priorities over time to fulfill orders within promised times, ensuring both high-priority orders and long-waiting low-priority orders receive timely attention. We present stochastic models of dynamic priority queueing networks to describe warehouse systems and estimate throughput times. Experiments validate the analytical stochastic models, and experimental results indicate that the dynamic priority model achieves shorter delay times than the static priority model and the FCFS model. We propose design insights based on experimental results and provide an approach to select the optimal robot number. Furthermore, by employing a fairness index, we develop a new decision support tool for determining warehouse configurations with requested performance objectives. Experimental results demonstrate that dynamic priority can ensure fairness across a wider range of scenarios. Additionally, with insufficient pickers, the system performs better with the put wall than without it.